batchin-public

# BatchIn

> Verification-first AI inference developer platform.

This repository contains public SDKs, a CLI, an MCP connector, and VaaS receipt verification clients. The private BatchIn control plane owns provider routing, model availability, pricing, billing, payment readiness, and customer data.

## Runtime truth

The authenticated API catalog at `https://api.batchin.tech/v1/models` is the only source of current model availability. Use a model only when the response reports verified provider smoke, pricing, usage, billing, and account entitlement. A model name, static README, test fixture, or HTTP 200 response is not evidence that inference will run.

The same evidence boundary applies to multimodal tasks, Stripe, x402, USDC, VaaS chain anchoring, and dedicated capacity. Check the relevant readiness and ledger endpoints before describing a capability as enabled or settled.

## Packages

- `@batchin/sdk` (TypeScript): authenticated requests, streaming, model discovery, and usage metadata.
- `@batchin/ai-sdk` (Vercel AI SDK): provider adapter.
- `batchin` (Python): authenticated client and framework adapters.
- `@batchin/cli`: authenticated API checks and model discovery.
- `batchin-mcp-server`: catalog, inference, pricing, trace, and receipt workflows.
- `@batchin/vaas` and `batchin-vaas`: receipt and evidence verification.
- `@batchin/contracts`: public Solidity interfaces; verify deployment status before use.

## Install

```bash
npm install @batchin/sdk
pip install batchin
```

See [README.md](README.md), package READMEs, and the [OpenAPI reference](https://api.batchin.tech/openapi.json) for request contracts. Package tests use deterministic fixtures and mocked transports and do not prove live provider or payment readiness.